This reverts commit fb21f3cc.

Add .gitignore for typical auxiliary files on Windows, renamed projects to match VS version, added x64 to VS2010.

git-svn-id: https://gflags.googlecode.com/svn/trunk@84 6586e3c6-dcc4-952a-343f-ff74eb82781d

git-svn-id: https://gflags.googlecode.com/svn/trunk@83 6586e3c6-dcc4-952a-343f-ff74eb82781d

git-svn-id: https://gflags.googlecode.com/svn/trunk@82 6586e3c6-dcc4-952a-343f-ff74eb82781d

git-svn-id: https://gflags.googlecode.com/svn/trunk@81 6586e3c6-dcc4-952a-343f-ff74eb82781d

Fix deprecation warnings of VC++ regarding strdup and unlink. Add solution files for VS 2010 and VS 2012.

git-svn-id: https://gflags.googlecode.com/svn/trunk@80 6586e3c6-dcc4-952a-343f-ff74eb82781d

git-svn-id: https://gflags.googlecode.com/svn/trunk@79 6586e3c6-dcc4-952a-343f-ff74eb82781d

Use project name (gflags) within include guards to avoid conflict with gperftools as reported by chen3feng (issue #62, http://code.google.com/p/gflags/issues/detail?id=62).

git-svn-id: https://gflags.googlecode.com/svn/trunk@78 6586e3c6-dcc4-952a-343f-ff74eb82781d

Fix "invalid suffix on literal; C++11 requires a space between literal and identifier" errors as reported by http://code.google.com/p/gflags/issues/detail?id=54.

git-svn-id: https://gflags.googlecode.com/svn/trunk@77 6586e3c6-dcc4-952a-343f-ff74eb82781d

git-svn-id: https://gflags.googlecode.com/svn/trunk@76 6586e3c6-dcc4-952a-343f-ff74eb82781d

	* gflags: version 2.0
	* Changed the 'official' gflags email in setup.py/etc
	* Renamed google-gflags.sln to gflags.sln
	* Changed copyright text to reflect Google's relinquished ownership


git-svn-id: https://gflags.googlecode.com/svn/trunk@74 6586e3c6-dcc4-952a-343f-ff74eb82781d

	* google-gflags: version 1.7
	* Add CommandLineFlagInfo::flag_ptr pointing to current storage (musji)
	* PORTING: flush after writing to stderr, needed on cygwin
	* PORTING: Clean up the GFLAGS_DLL_DECL stuff better
	* Fix a bug in StringPrintf() that affected large strings (csilvers)
	* Die at configure-time when g++ isn't installed


git-svn-id: https://gflags.googlecode.com/svn/trunk@72 6586e3c6-dcc4-952a-343f-ff74eb82781d

Clarify thread-safety comment for FlagSaver.

Tested:
blaze test base:commandlineflags_unittest

R=csilvers
DELTA=4  (3 added, 0 deleted, 1 changed)


Revision created by MOE tool push_codebase.
MOE_MIGRATION=3791


git-svn-id: https://gflags.googlecode.com/svn/trunk@71 6586e3c6-dcc4-952a-343f-ff74eb82781d

The last bit of rolling back commandlineflag categories:
remove the 'categories' field from CommandLineFlagInfo.  (Note
the code to fill this field was removed from
FillComandLineFlagInfo previously, so it's been an empty
string for some time now.)

R=ncalvin
DELTA=1  (0 added, 1 deleted, 0 changed)


Revision created by MOE tool push_codebase.
MOE_MIGRATION=3616


git-svn-id: https://gflags.googlecode.com/svn/trunk@70 6586e3c6-dcc4-952a-343f-ff74eb82781d

Get rid of the old FlagRegisterer constructor.

R=jkline,georgevdd
DELTA=27  (0 added, 27 deleted, 0 changed)


Revision created by MOE tool push_codebase.
MOE_MIGRATION=3609


git-svn-id: https://gflags.googlecode.com/svn/trunk@69 6586e3c6-dcc4-952a-343f-ff74eb82781d

Starting the rollback of flag categories.

I left in the old FlagRegisterer constructor.
I also left in 'categories' in CommandLineFlagInfo for now,
though I never use it.  I doubt anyone else does either, but I
want to minimize the number of ways this rollback can break
the build.  I will remove it in a subsequent CL.

R=ncalvin
DELTA=121  (28 added, 55 deleted, 38 changed)


Revision created by MOE tool push_codebase.
MOE_MIGRATION=3574


git-svn-id: https://gflags.googlecode.com/svn/trunk@68 6586e3c6-dcc4-952a-343f-ff74eb82781d

Add a 'flag_ptr' field to CommandLineFlagInfo that points to the current storage of the flag (i.e. &FLAGS_foo).

R=csilvers
DELTA=15  (15 added, 0 deleted, 0 changed)


Revision created by MOE tool push_codebase.
MOE_MIGRATION=3301


git-svn-id: https://gflags.googlecode.com/svn/trunk@67 6586e3c6-dcc4-952a-343f-ff74eb82781d

A user reported a bug that configure should die when g++ isn't
installed, but wasn't:
http://code.google.com/p/google-perftools/issues/detail?id=217

This turned out to be a bug in autoconf:
http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=357378

The workaround is simple: check for CXX before checking for
CC.  This means that if g++ is installed but gcc isn't, we
won't die (at cc-checking time), but I tested, and the
configure script dies later.  In any case, it seems unlikely
someone would have a c++ compiler installed but not a c
compiler.

This fixes the 4 opensource projects I work on that are
susceptible to this.

R=iant
DELTA=1437  (694 added, 694 deleted, 49 changed)


Revision created by MOE tool push_codebase.
MOE_MIGRATION=3222


git-svn-id: https://gflags.googlecode.com/svn/trunk@66 6586e3c6-dcc4-952a-343f-ff74eb82781d

Now I can get rid of the old constructor!

R=jkline,ncalvin
DELTA=28  (0 added, 27 deleted, 1 changed)


Revision created by MOE tool push_codebase.
MOE_MIGRATION=3169


git-svn-id: https://gflags.googlecode.com/svn/trunk@65 6586e3c6-dcc4-952a-343f-ff74eb82781d

http://code.google.com/p/google-gflags/issues/detail?id=45
reports that the error isn't always getting flushed on
cygwin.  So do that explicitly.

R=desovski
DELTA=1  (1 added, 0 deleted, 0 changed)


Revision created by MOE tool push_codebase.
MOE_MIGRATION=3140


git-svn-id: https://gflags.googlecode.com/svn/trunk@64 6586e3c6-dcc4-952a-343f-ff74eb82781d

Report category names via the xml help output (but not, yet,
during 'normal' help output).

R=ncalvin
DELTA=3  (3 added, 0 deleted, 0 changed)


Revision created by MOE tool push_codebase.
MOE_MIGRATION=3066


git-svn-id: https://gflags.googlecode.com/svn/trunk@63 6586e3c6-dcc4-952a-343f-ff74eb82781d

Add back in a backwards-compatible constructor for
FlagRegisterer.

Because this backwards-compatible API is intended to be
short-lived, I did it in the simplest, least invasive way
possible, which involved cutting-and-pasting.

R=ncalvin,jkline
DELTA=27  (27 added, 0 deleted, 0 changed)


Revision created by MOE tool push_codebase.
MOE_MIGRATION=3065


git-svn-id: https://gflags.googlecode.com/svn/trunk@62 6586e3c6-dcc4-952a-343f-ff74eb82781d

NOTE: I'm checking this in just to have a record in source control of the idea.  We've decided for now it doesn't make sense to go forward with flag categories -- and the __VA_ARGS__ for macros has caused lots of problems with uncaught accidental commas, already -- so a future commit will back this out.  (Actually, it's all a series of commits, for annoying technical reasons.)

Add support for flag categories.

In this CL, all you can do is set categories in the DEFINE_*
macros and then retrieve them via GetCommandLineFlagInfo and
similar.

In future CLs, we will start to give some semantic meaning to
particular flag values, as described in the designdoc.  In
particular, we will start to use flag categories to revamp
--help output.

Implementation-wise: to keep categories an optional macro
argument, I had to use __VA_ARGS__, which means future gflags
releases will no longer work with MSVC 7.1.  We're at MSVC 10
now, so I'm pretty much ok with that.

The downside of __VA_ARGS__ is there is no error if you
specify more args after the ones we expect.  To get around
that, I only use __VA_ARGS_ in this idiom:
static const OptionalDefineArgs var = { __VA_ARGS__ };
The new OptionalDefineArgs struct defines all the args that
may be optionally specified in the DEFINE_* macros.  For now,
that's only the 'categories' arg, though in theory more could be
added later.

R=titus,ncalvin
DELTA=92  (54 added, 3 deleted, 35 changed)


Revision created by MOE tool push_codebase.
MOE_MIGRATION=3057


git-svn-id: https://gflags.googlecode.com/svn/trunk@61 6586e3c6-dcc4-952a-343f-ff74eb82781d

Clean up the 'GFLAGS_DLL_DECL' stuff better: reduce use outside of windows/ directory.

R=jyrki
DELTA=33  (0 added, 27 deleted, 6 changed)


Revision created by MOE tool push_codebase.
MOE_MIGRATION=3016


git-svn-id: https://gflags.googlecode.com/svn/trunk@60 6586e3c6-dcc4-952a-343f-ff74eb82781d

Fix a bug in StringPrintf() that caused it to not print large strings properly.

R=nilton
DELTA=1  (0 added, 0 deleted, 1 changed)


Revision created by MOE tool push_codebase.
MOE_MIGRATION=2886


git-svn-id: https://gflags.googlecode.com/svn/trunk@59 6586e3c6-dcc4-952a-343f-ff74eb82781d

Add a new unittest case, that tests the --help output for a
really long helpstring.  Opensource gflags had a bug where we
were cutting off the output too soon; this test should protect
against such a thing.

R=nilton
DELTA=16  (16 added, 0 deleted, 0 changed)


Revision created by MOE tool push_codebase.
MOE_MIGRATION=2885


git-svn-id: https://gflags.googlecode.com/svn/trunk@58 6586e3c6-dcc4-952a-343f-ff74eb82781d

Ready to release gflags 1.6!

* google-gflags: version 1.6
* BUGFIX: Fix a bug where we were leaving out a required $(top_srcdir)
* Fix definition of clstring (jyrki)
* Split up flag declares into its own file (jyrki)
* Add --version support (csilvers)
* Update teh README for gflags with static libs
* Update acx_pthread.m4 for nostdlib
* Change REparseCommandLineFlags to return void (csilvers)
* Some doc typofixes and example augmentation (various)

R=jperkins
DELTA=44  (41 added, 0 deleted, 3 changed)


Revision created by MOE tool push_codebase.
MOE_MIGRATION=2838


git-svn-id: https://gflags.googlecode.com/svn/trunk@56 6586e3c6-dcc4-952a-343f-ff74eb82781d

It turns out we don't want all this dll-decl stuff for mingw.
So make the #ifdef correct for mscv-but-not-mingw.

I think the right fix would be to take all this logic out and
insert it via a MOE directive, just for the windows versions
of these files.  This is what ctemplate does.  But it's a
more-major change, and I'm looking to get this release out
today, so I'm submitting this (TBR), since it's a much more
minor change.


Revision created by MOE tool push_codebase.
MOE_MIGRATION=2837


git-svn-id: https://gflags.googlecode.com/svn/trunk@55 6586e3c6-dcc4-952a-343f-ff74eb82781d

mingw fixes:

1) mingw needs an #include to have access to mkdir.

2) It needs to always #include port.h (this is an identical
bit of code, in configure.ac, that I have in other opensource
projects for mingw support.)

3) I moved some code from port.cc to port.h, so I didn't have
to add logic to link in port.cc for mingw.

Last change before new release!  (*knock on wood*)  Submitting
TBR so I can get the release out today.  This isn't exactly a
trivial change, so I'm chary to submit TBR, but it's pretty
isolated to windows and mingw, and I've tested on those
platforms to make sure they compile and all tests pass.

DELTA=70  (37 added, 30 deleted, 3 changed)


Revision created by MOE tool push_codebase.
MOE_MIGRATION=2823


git-svn-id: https://gflags.googlecode.com/svn/trunk@54 6586e3c6-dcc4-952a-343f-ff74eb82781d

version of autotools (taht we're using now).  Get rid of them to clean
up the svn tree.  Also get rid of the 'python' file -- I don't know
what that is, but it's been a long time since the python version of
gflags has moved to its own project.


git-svn-id: https://gflags.googlecode.com/svn/trunk@53 6586e3c6-dcc4-952a-343f-ff74eb82781d

maintaining this opensource tree.  Some of this reorganization is
entirely a result of reorganization: blank lines in a few different
places, etc.  Here are the others:

1) I've added a new file, util.h, with lots of new, helpful routines,
most notably StringPrintf (printf returning a string).  I've also
moved some routines from the .cc and unittest.cc file here, such as
the CHECK macros, testing framework, and #ifdefs for 16-bit ints.
Quite a bit of code was rewritten to use these new routines.

2) I noticed that the special-case setenv() code was only needed for
windows systems, so I moved it to port.h/cc.

3) I also had to add a new vsnprintf wrapper in port.h/cc, to support
StringPrintf.

4) A few places I used an old name, commandlineflags, instead of
gflags.  Most or all of these should be fixed now.

5) Some of my copyright dates weren't quite right.  Fixed them up.

6) In some .cc files, I added using directives to not have to use
std:: so much.

7) I've added a minor new test, adding 10000 or so flags to see how
the system scales.

8) Some compile-warning fixes, such as int -> size_t when appropriate,
and better protected #defines in mutex.h

9) The .h files gained some logic defining GFLAGS_DLL_DECL.  This is
true even for .h files outside the windows directory, which will never
have these dll issues.  But one big advantage of my new organization
is auto-generating the windows versions of these files from the unix
versions, so there's some unnecessary (but harmless) duplication as a
result.

10) Fixed a bug in rpm.sh which would cause an unnecessary crash when
dpkg was missing.


git-svn-id: https://gflags.googlecode.com/svn/trunk@52 6586e3c6-dcc4-952a-343f-ff74eb82781d

(from 2.65).  I'm moving to a new system for maintaining the
opensource version, which involves using a 'hermetic' version of
autotools, and those are the versions installed in the hermetic
system.  There shouldn't be a user-noticeable change here.


git-svn-id: https://gflags.googlecode.com/svn/trunk@51 6586e3c6-dcc4-952a-343f-ff74eb82781d

	* Fix a bug where we were leaving out a required $(top_srcdir)


git-svn-id: https://gflags.googlecode.com/svn/trunk@50 6586e3c6-dcc4-952a-343f-ff74eb82781d

git-svn-id: https://gflags.googlecode.com/svn/trunk@49 6586e3c6-dcc4-952a-343f-ff74eb82781d

	* Add --version support (csilvers)
	* Fix doc typo (sudit)
	* Update the README instructions for gflags with static libs (csilvers)


git-svn-id: https://gflags.googlecode.com/svn/trunk@48 6586e3c6-dcc4-952a-343f-ff74eb82781d